Originally Posted by
angelozip
Neil was of an invaluable utility your comparative test: now we are sure that there is a loss of precision in the 3790 converted and which seems to have little to do with the ability of system "learning" and fine tuning itself, because they seem sufficiently used. Despite of your unquestionable conclusion I was even more curious about the causes of these discrepancies. Several could be the causes, but by reading the text indicated by Catymag (post 758) of Garmin engineering, seems that the key element in the calculation of accuracy of gps is the algorithm that calculates the deviation/error of the oscillator by introducing a correction factor (constant) based on the current internal temperature of the device. The logic indicates that the algorithm (I assume garmin's property) does not change between models tested but most likely only the correction factor. This constant (correction factor) must be specifies for each model, because it depends on the internal ambient temperature of each one, that has to do with processor heat dissipation, size of the case and other different items of hardware.
The disneysw's hint in post 755, what the variable table resulting of correction calculation is installed along with the firmware did not convince me because it does not explain several situations, as you even reported in earlier posts, that is, that there was no change in accuracy when you reinstalled the firmware original in your 3790, meaning that was not installed new table. By the way, this standard table is only installed when the unit leaves the factory, which may explain the great inaccuracy when installed the 3490 firmware on a new and unused 3790, as was in my case (but it was self adjusting with time of use). Remains the question of its location and how to reset this table, but that's another question. How does not seem to be the algorithm and nor the table that are installed with the firmware, another real candidate is the correction factor, specific for the model, which influences on calculation and can generate inaccuracy such as those observed in the comparison that you did, this is my hypothesis: the "cause" is this constant that which may be different for each model, that seems to be the case.
I believe that to have more security would be important someone to do a quick test, same as static, between a original 3790 (preferably never modified) and a 3790/3490, but with the original firmware (of 3790). The result we can deduce a few things. We already know by your own report that when the original firmware was reinstalled the on the 3790 modified, there was no change in accuracy compared to 3490 firmware, but the question is what will be the behavior of 3790 (mod but with original firmware) in comparison with the original 3790 never modified. If the result is not similar in accuracy we have problems that will not easy to solve. In this case would have the table variable that remained hidden in some file, database or fimware region changing the correction values and hence the accuracy? Correction factor would have been permanently recorded? Among other possibilities ... If the result is the same in both (original never changed and 3790 now with original firmware), it may mean that was installed new algorithm, new factor and/or new table by the firmware and everything needed for the result we should expect ..
If I failed at some, or all (lol), reasoning, correct me ..
Bookmarks